Controlling Spurious Oscillations in High-Order Fluid Simulations

66 concepts63 papers1 proposal

When simulating fluid flows, computational physicists face a fundamental mathematical conflict. To capture intricate features like turbulent eddies or acoustic waves efficiently, high-order numerical methods—such as Discontinuous Galerkin (DG) or high-order finite volume schemes—are highly desirable. By using high-degree polynomials to represent the fluid state, these methods achieve rapid error reduction as grid resolution increases.

However, this mathematical elegance breaks down when the fluid develops sharp gradients or shock waves. When a high-degree polynomial attempts to fit a sharp step function, it inevitably produces non-physical ripples at the interfaces. This behavior is known as the Gibbs phenomenon. In smooth regions, high-order methods excel; at discontinuities, they generate spurious numerical oscillations that can destabilize the entire simulation.

Why Numerical Ripples Break the Physics

To understand why these oscillations are catastrophic, we must look at the physical constraints governing fluid dynamics. Equations describing compressible flow, magnetohydrodynamics (MHD), or relativistic hydrodynamics assume that certain variables remain within strict physical bounds. For instance, mass density and fluid pressure must always be strictly positive.

When the Gibbs phenomenon introduces spurious oscillations near a shock wave, the numerical solution can overshoot or undershoot. In high-gradient regions, a sharp undershoot can easily drop below zero, yielding a negative pressure or density. As Sibusiso Mabuza et al. observe in their work on ideal MHD equations, such violations of fundamental physical constraints lead to immediate simulation failure.

Mathematically, a negative pressure causes the local speed of sound ($c = \sqrt{\gamma P / \rho}$) to become an imaginary number. This instantly destroys the hyperbolic nature of the governing partial differential equations, turning a well-behaved wave propagation problem into an unstable, chaotic system that causes the code to crash. This fragility is particularly severe in extreme regimes. For example, Forrest W. Glines et al. highlight that high-order simulations of relativistic hydrodynamics are highly prone to generating unphysical states or superluminal velocities near sharp discontinuities.

Furthermore, these oscillations are not confined to physical space. When propagating fluid simulations under uncertainty, Alina Chertock et al. demonstrate that the Gibbs phenomenon also manifests within the stochastic domain, distorting statistical metrics like mean and variance.

Standard High-Order Polynomial Approximation:
       ▲
       │       ┌─┐   ┌─┐
State  │      ┌┘ └┐ ┌┘ └┐  ◄── Spurious Gibbs Oscillations
       │  ────┘   └─┘   └─────────────
       │        ▲
       │        └─ Discontinuity (Shock)
       └──────────────────────────────► Space

The Mechanism of Subcell Limiting

To prevent these crashes without destroying the accuracy of the high-order method in smooth regions, modern solvers employ subcell limiting.

Historically, shock-capturing was achieved by adaptively shifting the interpolation stencils away from discontinuities, a concept pioneered by Chi‐Wang Shu via Essentially Non-Oscillatory (ENO) schemes. The global stability of this mathematical framework was later rigorously proven by Ulrik S. Fjordholm et al.

Modern element-based methods, such as the Discontinuous Galerkin Spectral Element Method (DGSEM) or Flux Reconstruction (FR), operate within fixed computational cells. Dropping the polynomial degree of an entire cell down to a first-order approximation when a shock appears is highly wasteful and dampens fine-scale features nearby. Subcell limiting solves this by zooming inside the troubled element.

High-Order Element with Subcell Limiting:
┌─────────────────────────────────────────┐
│  Smooth Subcells       │ Shock Subcell  │
│  (High-Order Polynomial)│ (Low-Order FV) │
│        ~ ~ ~ ~         │     ───        │
└─────────────────────────────────────────┘

When a cell is flagged as containing a discontinuity, it is internally subdivided into a fine subcell grid.

  1. In subcells where the flow remains smooth, the solver retains the high-order polynomial representation.
  2. In subcells directly intersecting the shock, the high-order operator is locally deactivated and blended with a highly robust, low-order, oscillation-free method—typically a first-order Finite Volume (FV) scheme.

By confining the low-order numerical dissipation strictly to the subcells that contain the discontinuity, the scheme stabilizes the shock interface while preserving high-order fidelity everywhere else. Huajun Zhu et al. successfully implement this subcell blending strategy within Gauss-point DGSEM frameworks, while Arpit Babbar et al. apply a similar subcell blending limiter to Multi-Derivative Runge-Kutta Flux Reconstruction schemes to guarantee physical admissibility near strong shocks.

Enforcing Physical Constraints and Entropy Stability

While limiters suppress oscillations by altering the local spatial representation, another powerful approach enforces physical consistency directly through mathematical invariants: Physical Constraint Preservation (PCP) and Entropy Stability.

PCP methods act as a nonintrusive mathematical safety net. Instead of altering the underlying solver, they apply optimization-based filters to the computed polynomial states. For instance, Mike Kirby et al. developed structure-preserving nonlinear filters using convex optimization. If a high-order method produces a state that dips into negative territory, the filter mathematically projects the polynomial coefficients back into the valid, positive invariant domain while strictly maintaining mass and momentum conservation. This approach is vital in extreme environments, such as the high-Mach astrophysical simulations designed by Dinshaw S. Balsara et al., or the relativistic regimes explored by Shankar Basak et al.

Complementing this is the framework of entropy stability. In continuous physics, fluid flows must obey the second law of thermodynamics, meaning mathematical entropy must always be dissipated across a shock wave. Standard high-order schemes do not automatically satisfy this property at the discrete level, which can cause errors to accumulate into unphysical artifacts like the "carbuncle phenomenon"—where a shock wave develops jagged, grid-aligned protrusions.

To fix this, researchers design numerical fluxes that mimic continuous entropy bounds. As demonstrated by Praveen Chandrashekar, constructing schemes that are simultaneously kinetic energy preserving and entropy stable prevents high-frequency oscillations from growing exponentially. By extending entropy stability to complex geometries and varying-width channels, Jesse Chan et al. show that ensuring discrete thermodynamic consistency provides a rigorous mathematical guarantee against simulation failure, allowing high-order methods to remain stable in the presence of severe transonic gradients.

Key Foundational Work

Efficient implementation of essentially non-oscillatory shock-capturing schemes (1988) — contributed 1 concept elements to this direction.
Accurate computations of compressible Navier-Stokes equations by the spectral collocation method. (1990) — contributed 1 concept elements to this direction.
History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_gibbs_phenomenon_spurious_numerical_osci_fb1bc6d4,
  title = {Gibbs phenomenon (spurious numerical oscillations)},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/gibbs-phenomenon-spurious-numerical-oscillations/rdr_fb1bc6d48ed8483e8152c43e3a954035},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-19},
}

This exact version (version 1, frozen text):

@misc{lacuna_gibbs_phenomenon_spurious_numerical_osci_fb1bc6d4_v1,
  title = {Gibbs phenomenon (spurious numerical oscillations)},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/gibbs-phenomenon-spurious-numerical-oscillations/rdr_fb1bc6d48ed8483e8152c43e3a954035/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-19},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake